Hacker News Reader

Comment by tlrobinson | Hacker News Reader

While I think using password managers with random passwords is far better than sharing the same password between every account, I've never really gotten comfortable with storing passwords in a file on my computer.

What I'd really like is a password manager hardware dongle of some kind, like the Bitcoin Trezor wallet.

View on HN

It doesn't matter. If you have malware lurking in your computer it will just snarf your passwords from the wire and then you're owned all the same. If you use some sort of auth signing system, the request can just be intercepted and modified on the fly.[0] The Trezor is next to useless even for bitcoin for this very reason. Sure they can't steal your money directly, but just replacing the addresses you see and send to accomplishes exactly the same thing. If your platform isn't trusted, no amount if smart crypto or hardware dongles can make it safe.[1]

[0]: The rebuttal for this will be signing every request with details of it with a hardware dongle, but would you want to do this for every action in your email client? If the answer is "no", you're owned. Ultimate security is unusable, and doesn't really solve anything outside of the most astute of professional users. Just writing this post I would need 4 signatures, one to log in, one to post, one to fix a typo, and yet another to add this footnote. Would I be able to handle that? No way, I'm far too lazy for that.

[1]: The issue is that perfect compromise is impossible to detect. I can be reasonably confident on a heavily sandboxed device like an iPhone that there's little in the way of malware that would affect me. The downside being that I have no tools or methods of analysis if I thought it was compromised. There's no such confidence on the computers I use on a daily basis. I've always thought we have confirmation bias with malicious software. We only notice the dumb stuff while the smart goes unnoticed.

For passwords you're right, but at least it prevents theft of every password all at once.

For Bitcoin hardware wallets like Trezor, IIRC they either do or will support BIP-70 "Payment Protocol" payment requests that are signed with an X.509 cert, allowing you to verify the request on the dongle's screen.

Nothing stops a threat from just lying and waiting for you to expose a large number of passwords. Having one stolen doesn't raise red flags in itself.

I don't think signed addresses will be particularly effective. With the sort of key stores we have now, it seems pretty plausible that a bad actor to get a certificate that would pass on the Trezor device. It raises the barrier of entry a little though.

It should be just as effective as HTTPS is. Of course whether that's effective enough depends on who you ask. Yes, rogue/compromised CAs are occasionally a problem, but it generally works pretty well.

It doesn't hold up to SSL stripping very well. As we are working under the assumption of a compromised host, the absence of a signature on the Trezor when you don't expect one wouldn't raise any suspicion. The omnipotent host malware can remove all references to the payment request being signed from the payment gateway before the user sees it.

Doesn't the TREZOR show you the address you're sending from/to as well as the amounts? That would be pretty easy to double-check.

Double check it against what information source? The malicious software has altered the one shown on both devices. They will of course match, but not with the one you're actually meant to be paying to.[0][1]

[0]: http://www.reddit.com/r/Bitcoin/comments/23sjle/chrome_exten....

[1]: http://www.reddit.com/r/Bitcoin/comments/1vrium/a_google_chr...

This is eventually going to be solved to a large degree by things like FIDO Universal 2-Factor dongles or whatever they evolve into over the next few years.

It'll basically push specialized requirements to the hardware dongle (ie deciding whether it's enough to confirm user registration/authorization with the touch of a buttom, or whether it needs to be with a 4 digit pin, or even with biometrics like voice or fingerprint or iris scan). The test device in the following presentation video only uses a button press to confirm user intent, but it could have arbitrary requirements, making the protocol usable for both trivial website logins to online banking to eventually perhaps even a replacement for defense department CACs.

Here's a good fairly in-depth video presentation of what the FIDO Alliance is working on (there are functional test devices, and a functional test branch of Chrome that works with them):

http://www.yubico.com/2014/07/future-online-authentication/

This is the future we all need, and if the timelines are correct, it will be released this year, so very soon. Note that Apple should be releasing NFC soon, so it would probably work with this. Everyone wants to stop phishing, as Apple knows very well, and this is the solution. Apple will hopefully jump on board after what just happened. Nudge, nudge.

If the dongle sends cleartext passwords to the computer then the passwords may still be captured. If the dongle sends one-time passwords then it's basically a Yubikey.

I find myself suggesting this to people often, but I'm excited about the Mooltipass[0], an offline password keeper crowd developed on Hackaday. Hope to get one when they've ironed out the bugs and finally move to random passwords.

[0] http://hackaday.io/project/86-Mooltipass

Reply on news.ycombinator.com